Falsehoods programmers believe about languages (localization)

I would add: a user's country of residency can be mapped 1-to-1 to the user's preferred language, or at least to a language the user understands.

So, SO many websites get this completely wrong.

Back when I was making WIN32 apps (that at the time didn't have a layout manager), the hardest languages to localize for were:

* German, simply because the German translations we were using were twice as long. I don't speak German so I couldn't judge if all those extra words were required.

* Simplified Chinese, because the default font in that language was huge - single characters would take twice the width (and more irritatingly, height) of an English character.

* Japanese, partly because the default font was also large but mostly because the multibyte character codes (this was before unicode support) for common words included bytes that were commonly used for separators in other languages. So things like backslash path seperators (which Japan didn't use) would have to be handled with extreme care in the code.

Sorting should probably have its own list.

"Ordering of words follows simply from ordering on the letters in the alphabet" - nope, Spanish "ll" is not in the alphabet and is sorted like a single special letter.

"It's always lexicographical left to right or at least lexicographical in a single, consistent direction?" - nope, accents in French (è,é) are compared right-to-left.

Pretty much the only thing I could find in the whole list that I thought anyone might believe is "every language has words for yes and no". But here's another myth that I find it hard to understand how anyone could believe: "interface design decisions are generally made by programmers". If a site uses country flags to indicate languages, it probably wasn't a coder that decided to do that, and whoever did make the call didn;t believe there was a one-to-one relationship between countries and languages. I particular, pretty much everyone can think of at least one country other than the USA where they speak English.

If you want to create a good piece of software (let's say Photoshop) in English, create it in English. If you want to create a good Photoshop in German, create it again but in German. You can use the English one as a base, but treat it as a new product that happens to have a lot of code reuse, not a variant of English Photoshop.

Then, to reduce maintenance costs, you can fold the changes back in to the original software, using conditional statements and so on. This creates International Photoshop, of which English Photoshop, German Photoshop and Chinese Photoshop are just configuration options.

What you should not do is create International Photoshop first, and then specialize it for English and German and Chinese. You'll find places where the localizations don't fit into the framework you created, and then you'll either go back and change the framework, or (more likely) you'll just use English conventions in the non-English versions to make them fit the framework, and people in other countries will just live with with their language's equivalent of Engrish.

This is related to "a library must have 3 clients", "make a game not a game engine", etc, also "red green refactor". You have to make several specific things before you can generalize anything. If you don't make the concrete things first, you don't know what has to be generalized and how, and what doesn't. When you make the framework first, it has all the wrong generalization points.

Of course, if you already did this for a previous project so you already know how the framework should be, then you can try the framework-first approach. But beware - with that approach you might get stuck in the second system effect.

This comment has been delayed for hours by HN's rate limit.

I can easily add another one, on change conversion:

Changing case of a word can be performed by changing case of each letter individually.

This is a good list, but it needs examples.

Another one: depending on the language, 0 is either singular or plural. In French it's singular and it's always fun to see the denial in the eyes of my French colleagues when I tell them that 0 is plural in English.

Good list, but I imagine most programmers don't believe most of these falsehoods and are aware of a lot of these issues.

It's just not always easy to deal with them correctly - and quicker/easier/cheaper to just do the thing that will be right for the large majority of your users.

Number 1 falsehood I’ve seen is: Language = Locale

I came across one of this cases in Python where .title() capitalizes every word of the sentence, but in Spanish you only capitalize the first letter.